Ask what makes a roof suitable for solar and the first answer is almost always that it needs to face south. It is the one piece of the subject that has reached general circulation, and it is wrong by a wide margin in the direction that costs people money: roofs get ruled out that would have been fine.
The figures below are modelled in PVGIS for a 4 kWp crystalline silicon array, roof-mounted, in the English Midlands, holding everything constant except the direction it points.
What direction costs
| Orientation | Annual output (kWh) | Against due south |
|---|---|---|
| Due south | 3,761 | 0% |
| 15° east of south | 3,747 | −0.4% |
| 15° west of south | 3,726 | −0.9% |
| 30° east of south | 3,681 | −2.1% |
| 30° west of south | 3,640 | −3.2% |
| South-east | 3,571 | −5% |
| South-west | 3,514 | −6.6% |
| 60° east of south | 3,419 | −9.1% |
| 60° west of south | 3,350 | −10.9% |
| Due east | 3,029 | −19.4% |
| Due west | 2,955 | −21.4% |
| Due north | 2,072 | −44.9% |
Each row is the same array, same tilt, same location, turned to a different compass bearing. East beats west slightly here, which is a real feature of the climate record rather than a rounding artefact: British mornings are on average clearer than British afternoons.
The shape of that table is the point. The penalty is almost flat across the whole southern half of the compass and only becomes serious once you pass east or west.
- Within 15° of south, the loss is under 1%. That is inside the year-to-year weather variation and not worth discussing.
- At 30° off, it is 2.1% to the east and 3.2% to the west. A roof at 30° off south is, for practical purposes, a south-facing roof.
- At 45° off, between 5.0% and 6.6%. Still a perfectly good installation.
- Due east or due west costs around a fifth. Significant, and still frequently worth doing, for reasons that are about tariffs rather than physics.
- Due north loses 44.9% and should generally not be built.
Anyone who has been told their roof is unsuitable because it faces south-west is looking at a 3.2% penalty.
Tilt matters less than direction
The same exercise, holding the array due south and varying the pitch.
| Tilt | Annual output (kWh) | Against 35° |
|---|---|---|
| 0° | 3,146 | −16.3% |
| 10° | 3,414 | −9.2% |
| 20° | 3,611 | −4% |
| 25° | 3,682 | −2.1% |
| 30° | 3,731 | −0.8% |
| 35° | 3,761 | 0% |
| 40° | 3,770 | +0.2% |
| 45° | 3,759 | 0% |
| 50° | 3,729 | −0.8% |
| 60° | 3,607 | −4.1% |
| 90° | 2,774 | −26.2% |
Optimum sits near 40° at this latitude, but the curve is flat around it. The range 25° to 50° spans 2.1%.
British pitched roofs are generally built somewhere between 30° and 45°. That band sits inside one per cent of the optimum. Tilt is therefore almost never a reason to reconsider an installation on a pitched roof, and paying for tilt frames to correct it on such a roof is paying for nothing.
Flat roofs are a different case. Laying panels flat costs 16.3%, so frames genuinely earn their keep there. They also introduce row-to-row shading and wind loading, which is a design problem rather than an angle problem.
Total output is the wrong thing to optimise
Everything above ranks orientations by annual kilowatt-hours. That ranking only decides earnings if every kilowatt-hour is worth the same amount, and it is not. A unit you consume in the house saves you the retail price. A unit you export earns your export rate, which is a fraction of it.
So the orientation that generates most is not automatically the one that earns most. A south-facing array concentrates its output around midday, which for a household that is out all day is exactly when it cannot be used. An east-west split spreads the same total across morning and late afternoon, generating less overall but matching consumption better.
That trade is worked through in why an east-west roof can beat a south-facing one, and the underlying economics in why a unit you use is worth more than a unit you export.
Getting the number for your own roof
PVGIS is free, public, and covers Europe including the UK. Entering your own coordinates, pitch and bearing takes a couple of minutes and produces a figure with its own stated database and averaging period behind it.
Two cautions when you do. It models the terrain horizon from elevation data but knows nothing about chimneys or trees, so subtract whatever object shading applies. And its default system loss of 14% is an assumption, not a measurement of your installation.
Once you have a yield figure, what it is worth depends on your consumption and your export rate. Our calculator does that half, and reports the result as a range with the lower bound stated.
PVGIS 5.2 (European Commission Joint Research Centre), PVGIS-SARAH2 radiation database, 2005–2020 average. 4 kWp crystalline silicon, roof-mounted, 14% system loss, Birmingham, UK. Retrieved 2026-09-12. Source
Same array, same pitch, same place. Only the bearing changes. Select a petal.
Due south
3,761kWh/yr
The reference
- Within 30° of south
- under 3.5% lost
- Due east
- -19.4%
- Due west
- -21.4%
- Due north
- -44.9%
East beats west here. British mornings are on average clearer than British afternoons, so the shape is genuinely lopsided rather than drawn that way.
PVGIS 5.2 (European Commission JRC), PVGIS-SARAH2, 2005–2020. 4 kWp, roof-mounted, 14% system loss, Birmingham, UK.